Research And Prototype Implementation Of Content-based Video Retrieval

Content-based video retrieval system is a system which structures the video data and uses the still features and motion features to retrieval wanted videos by similarity degree.This paper does research on technique of video retrieval system, including structuring video, shot boundary detection, key-frame selection and feature extraction and matching algorithms. In order to make feature extraction standard, this paper does research on MPEG-7 standard's scope, application areas and all visual components. And further explores the camera motion and motion trajectory.In this paper, the scalable color, homogeneous texture and contour-based shape descriptors are united to describe the pictures. The extracting and matching algorithms are given in it. The experiments show that this method is robust. Motion Activity Descriptor was chosen as the motion feature to combine with the three still features to describe the video.A contented-based video retrieval system model was implemented in this paper. It's extraction of still and motion features are based on the MPEG-7 standard. And the system model has supports of shot database and feature database. It implemented features' extraction, putting them into databases, query-by-example and return of videos by similarity degree. Also it can browse videos by motion activity intensity.